On Feb 7, 2008, at 10:01 PM, Jan Wieck wrote:

> It not only can, it definitely will hurt performance. A Slony  
> replica experiences the same updates, row per row, as the origin  
> did. This means that the need for vacuuming is exactly the same.

Additionally, if your bloat makes your PK indexes bigger (and more  
sparse) then you also magnify the effect of updates that affect many  
rows.  On the master, it would be one query but on the replica it  
would be expanded to individual row updates.  Having to look up each  
PK in the index will be slowed down by the bloat.

I'd recommend taking a snapshot of your relpages numbers for each  
table and index from the pg_class table, running vacuum and then re- 
index on the big tables, then compare the relpages numbers again.   
Then compare performance and see how the bloat was affecting you.